Manufacturability

Manufacturability Rank

Manufacturability rank determined by the Manufacturability in silico assay (MISA) is a screen to predict the packaging efficiency of a given AAV gene therapy vector design by identifying and quantifying truncation hotspots. Here, a probabilistic method is used to assess the likelihood of hotspots by comparing the predictions of a new construct against a distribution of predictions from a training set. For each position in a sequence, the model makes a prediction on the truncation hotspots, and this is compared to the cumulative distribution of similar predictions from the training set. A probability is assigned based on where this prediction falls in the distribution. These probabilities are converted to negative log-likelihoods and summed across the construct. The sum of these negative log-likelihoods is the manufacturability rank. A higher score suggests more hotspots.

Comparison of NGS data and predicted truncation peaks. The figure displays the NGS data alongside the truncation peaks for a given construct (forward and reverse replication). X-axis represents the position along the sequence, and y-axis represents the read count. Red bars indicate actual NGS data, and blue lines represent model predictions. Areas where red and blue bars overlay suggest accurate predictions by the model. Discrepancies between the two may indicate regions where the model's predictions differ from the observed data.